AI Composer — Markov-Chain Melody Generator

Composition Style

Generator

Temperature (randomness) 1.00

Tempo 96 BPM

Voice

Live State

Do (1)

Notes generated

Entropy H(n)

0.00 bits

—

How it works

This is a first-order Markov-chain composer, the same idea behind Hiller & Isaacson's 1957 Illiac Suite — the first music composed by a computer. Eight scale degrees (a major scale, Do → Do′) are states of a Markov chain. The next note depends only on the current one, drawn from a learned transition matrix:

P(next = j | current = i) = row i of the transition matrix
Σ_j P(j | i) = 1   for every state i

Each style defines a different matrix: Classical weights nearby scale steps highest (a Gaussian falloff in scale-degree distance), Pentatonic Folk restricts the reachable degrees to a five-note subset, Jazz favours wider leaps (thirds and fourths), and Random Walk is uniform over all eight degrees.

Temperature reshapes the chosen row before sampling, exactly like softmax temperature in a language model:

p'(j) ∝ p(j)^(1/T)
T < 1 → sharper, more predictable (greedy)
T > 1 → flatter, more surprising (exploratory)

The live entropy readout is the Shannon entropy of that reshaped row, H = −Σ p'(j)·log₂p'(j), in bits — a direct measure of how unpredictable the AI's next choice is. On the lattice, curved arcs fan out from the currently active node with opacity proportional to probability; the glowing marker travels along the sampled arc and a tone sounds at the destination note, so you both see and hear the same probability distribution driving the composition.

Active note / marker Reachable next notes

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.